# Google Cloud Run

Official pricing: https://cloud.google.com/run/pricing  
Category: hyperscaler · Isolation: gvisor

## Pricing regimes (raw)

- **Services, instance-based billing ("CPU always allocated"), Tier 1** (resource): $0.0648/vCPU-h, $0.0072/GiB-h
- **Services, request-based billing (default), Tier 1** (resource): $0.0864/vCPU-h, $0.009/GiB-h
- **Services/Jobs, instance-based billing, second-gen execution environment (microVM VM sandbox), Tier 1** (resource): $0.0648/vCPU-h, $0.0072/GiB-h
- **Cloud Run CUD 1y/3y, instance-based** (resource): $0.053784/vCPU-h, $0.005976/GiB-h [commit]
- **Compute Flexible CUD 1y, instance-based** (resource): $0.046656/vCPU-h, $0.005184/GiB-h [commit]
- **Compute Flexible CUD 3y, instance-based** (resource): $0.034992/vCPU-h, $0.003888/GiB-h [commit]
- **Worker pools (no ingress, always-on background), Tier 1** (resource): $0.040478/vCPU-h, $0.004446/GiB-h
- **Worker pools + Flexible CUD 3y** (resource): $0.021859/vCPU-h, $0.002401/GiB-h [commit]

## Features

Yes: volumes, idle auto-stop, ≥24 h sessions, custom image (Docker or snapshot), your own Docker/OCI image, root, HTTPS preview URLs, custom domains, egress allowlist, open internet, static egress IP, private networking, SOC 2, HIPAA, SSO, GPU, Python, Node.js, wake on request, webhooks, audit logs, EU data residency, spend limits, MCP server, fast boot (benchmarked), gVisor or VM (no shared kernel), inbound access rules, extra volumes, shared volumes

No: snapshots, memory snapshots, fork/clone, pause/resume, persistent disk, start from your own snapshot, full VM (own kernel), Docker inside, nested virtualization, systemd, browser, desktop GUI, computer-use API, code interpreter, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, preinstalled agents, SSH, public IPv4, raw TCP inbound, self-hosting, BYOC, open source, arm64, Windows, macOS, GPU desktop, HTTP method/path egress rules, live resize, live fork (no pause), memory fork, automatic snapshots, snapshots on demand, agent harness API, hosted agent API (their own agent), firewall inside (nftables), secret proxy, secret proxy for any API

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/google-cloud-run.json → feature_evidence

## Caveats

- Rates from the official pricing page tables (raw/cloud_google_com_run_pricing.txt, us-central1 view). The Cloud Billing Catalog API needs an API key and was not used.
- Tier 2 (asia-east2, asia-northeast3, asia-southeast1/2, asia-south2, australia-southeast1/2, europe-central2, europe-west2/3/6/10/12, me-central1/2, northamerica-northeast1/2, southamerica-east1/west1, us-west2/3/4): instance-based/jobs $0.0000216/vCPU-s + $0.0000024/GiB-s (+20%); request-based active $0.0000336/vCPU-s + $0.0000035/GiB-s and min-instance idle $0.0000035/$0.0000035 (+40%); worker pools $0.000013493/$0.000001482; delayed jobs $0.00001512/$0.00000168; Instances $0.000000324/$0.000002316; ephemeral disk $0.000131507/GiB-h. Tier 2 GPU (asia-southeast1): L4 $0.00022404/s, RTX PRO 6000 $0.00043826/s. northamerica-south1 (Mexico), africa-south1, asia-northeast2, asia-southeast3/4, europe-north2, europe-west8, me-west1, us-east5, us-south1, us-west1, us-west8 are Tier 1. Cheapest EU/Asia regions are Tier 1 = US price, so region modes collapse into one.
- Free tier (per billing account, monthly, applied as a Tier 1 spend discount): instance-based services and jobs 240,000 vCPU-s + 450,000 GiB-s ($5.22 derived = free.monthly_credit); request-based 180,000 vCPU-s + 360,000 GiB-s + 2M requests ($6.02); delayed jobs 342,857 vCPU-s + 642,857 GiB-s; worker pools 384,204 vCPU-s + 728,744 GiB-s. $300 new-customer trial credit is generic GCP (not re-verified this session).
- Egress: Premium Tier only; 1 GiB/month free within North America, then $0.12/GiB (0-1 TiB) to NA/EU/Asia, $0.11 (1-10 TiB), $0.08 NA / $0.085 EU-Asia above 10 TiB; Australia/South America/Korea/Indonesia $0.19 then $0.18/$0.15; China $0.23. Same-region traffic to GCP services, Cloud CDN and Cloud Load Balancing is free.
- Static egress IP requires Serverless VPC Access or Direct VPC egress + Cloud NAT (NAT pricing not captured). No public IPv4 per instance.
- Ephemeral disk (optional) $0.000109589/GiB-h (~$0.08/GiB-month); flex CUD 1y $0.000078904, 3y $0.000059178. The default filesystem is in-memory and counts against RAM.
- GPU zonal-redundant surcharge: L4 $0.0002909/s ($1.047/h), RTX PRO 6000 $0.00056913/s ($2.049/h); modes use non-zonal rates. GPU services need >= 4 vCPU/16 GiB (L4) or 20 vCPU/80 GiB (RTX PRO 6000).
- The "Instances" table on the pricing page ($0.00000027/vCPU-s, $0.00000193/GiB-s) has no explanation on the page; not modelled.
- CUD modes are modelled as always-on commitments; delayed-jobs prices can change every 30 days.
- Budget spend caps (Preview, 2026-07-27): when a Cloud Billing spend cap is hit, Cloud Run pauses services (5xx), jobs and worker pools until the cap is lifted.
- Ephemeral disk is Preview (2026-04-20). GPU: must use instance-based billing; RTX PRO 6000 not offered in every Tier 1 region (e.g. europe-west1 lists L4 only). Pricing-page examples 5 and 6 ($0.45, $16.83 before free tier) are computed at Compute Flexible CUD 3y rates, not list.

## How this provider charges


Cloud Run bills allocated vCPU-seconds and GiB-seconds, rounded up to 100 ms, with different rates depending on the
**billing configuration** (instance-based vs request-based), the **resource type** (services, jobs, delayed jobs,
worker pools) and the **region tier**. Rates are from the official pricing tables at https://cloud.google.com/run/pricing
(page dump raw/cloud_google_com_run_pricing.txt, us-central1 view). The Cloud Billing Catalog API needs an API key and
was not used. Tier 1 regions include us-central1/us-east1/us-east4, europe-north1/west1/west4/southwest1/west9 and
asia-south1/northeast1/east1, so the cheapest EU and Asia regions cost the same as the US.

Reference 4 vCPU / 8 GiB: instance-based **$0.3168/h**; request-based active **$0.4176/h**; worker pool **$0.1975/h**.

## Regime table

| Regime | When it applies | How billed | Numbers (per vCPU-s / per GiB-s) | Source |
|---|---|---|---|---|
| Services, instance-based billing (formerly "CPU always allocated") | `billing: instance` | Whole instance lifetime, 1-minute minimum; no request fee | $0.000018 / $0.000002 ($0.0648 / $0.0072 per h) | pricing page |
| Services, request-based billing (default) | Scale-to-zero HTTP services | Only while starting, shutting down, or handling >= 1 request, 100 ms rounding; + requests | Active $0.000024 / $0.0000025; requests $0.40 per 1M | pricing page |
| Request-based idle min-instances | `min-instances > 0`, idle | Reduced idle rate; idle non-min instances are free | $0.0000025 / $0.0000025 | pricing page |
| Cloud Run CUD 1y or 3y | Spend commitment, Cloud Run only | Same rate for 1y and 3y | Instance-based $0.00001494 / $0.00000166 (-17%); request-based $0.00001992 / $0.000002075, requests $0.332/M | pricing page |
| Compute Flexible CUD 1y / 3y | Spend commitment shared with GCE/GKE | Discount on instance-based | 1y $0.00001296 / $0.00000144 (-28%); 3y $0.00000972 / $0.00000108 (-46%) | pricing page |
| Jobs | Run-to-completion tasks (up to 168 h; 1 h with GPU) | Instance-based rates, 1-minute minimum | $0.000018 / $0.000002 | pricing page |
| Delayed Jobs | Deferred-start jobs | Dynamic price, "can change up to once every 30 days" | $0.0000126 / $0.0000014 (-30%); flex CUD 1y $0.000009072 / $0.000001008, 3y $0.000006804 / $0.000000756 | pricing page |
| Worker pools | Always-on background workers, no ingress | Instance lifetime | $0.000011244 / $0.000001235 (= Fargate x86); flex CUD 1y $0.000008096 / $0.000000889, 3y $0.000006072 / $0.000000667 | pricing page |
| "Instances" table | Unexplained on the page | ? | $0.00000027 / $0.00000193 | pricing page |
| GPU | L4 / RTX PRO 6000 attached to services, jobs, pools | Per second on top of CPU/RAM | L4 $0.0001867/s ($0.672/h) non-zonal, $0.0002909/s zonal; RTX PRO 6000 $0.00036522/s ($1.315/h), zonal $0.00056913/s | pricing page |
| Ephemeral disk (optional) | Disk beyond the in-memory FS | Per GiB-hour | $0.000109589 (~$0.08/GiB-month); flex CUD 1y $0.000078904, 3y $0.000059178 | pricing page |
| Tier 2 regions | europe-west2/3/6/10/12, asia-southeast1/2, asia-east2, asia-northeast3, australia-*, northamerica-*, southamerica-*, us-west2/3/4, me-central* | Higher unit rates | **not captured** (page shows the us-central1 view) | pricing page region list |
| Free tier (monthly, per billing account, Tier 1-priced discount) | All accounts | Deducted | Instance-based/jobs 240,000 vCPU-s + 450,000 GiB-s ($5.22); request-based 180,000 vCPU-s + 360,000 GiB-s + 2M requests ($6.02); delayed jobs 342,857 + 642,857; worker pools 384,204 + 728,744 | pricing page |
| Internet egress (Premium Tier) | Outbound internet | Tiered per destination | 1 GiB/month free within North America; $0.12/GiB 0-1 TiB, $0.11 1-10 TiB, $0.08 (NA) / $0.085 (EU, Asia) above; Oceania/SA/Korea/Indonesia $0.19; China $0.23. Same-region GCP, Cloud CDN, Cloud LB: free | https://cloud.google.com/vpc/network-pricing |
| New-customer trial | New GCP accounts | Credits | $300 / 90 days (generic GCP, not re-verified) | cloud.google.com/free |

## Gotchas

1. **Request-based is 33% dearer per second but free between requests**; instance-based is cheaper only if the instance is busy most of its life.
2. **60-minute request timeout** on services: a sandbox session must be one long request or many short ones; jobs run up to 168 h but have no ingress.
3. **1-minute minimum** per instance on instance-based billing and jobs.
4. **In-memory filesystem**: files written to disk consume RAM you pay for unless you add the ephemeral disk.
5. **Egress is expensive**: $0.12/GiB with only 1 GiB free (vs AWS 100 GB free); Standard Tier isn't used by Cloud Run.
6. **Static egress IP needs VPC egress + Cloud NAT** (extra, not priced here).
7. Delayed-jobs prices float monthly; CUDs bill every hour of the term.

## Worked example

4 vCPU / 8 GiB, 50 concurrent x 8 h/day x 22 days = **8,800 instance-hours**, 30% CPU (no effect: allocated
billing), 50 GiB state (no snapshot feature; would need GCS/Filestore, not priced), 100 GiB egress
(1 GiB free, 99 x $0.12 = $11.88).

| Regime | Compute | Total / month |
|---|---|---|
| Instance-based service | 8,800 x 0.3168 = $2,787.84 - free $5.22 | **$2,794.50** |
| Request-based, request in flight 100% of session | 8,800 x 0.4176 = $3,674.88 - free $6.02 (requests negligible) | **$3,680.74** |
| Request-based, request in flight 30% of session | 2,640 x 0.4176 = $1,102.46 - $6.02 | **$1,108.32** |
| Worker pool (no inbound) | 8,800 x 0.19748 = $1,737.84 - free $5.22 | **$1,744.50** |
| Instance-based + flex CUD 3y sized for 50 x 24/7 | 50 x 730 x 0.171072 = $6,244.13 | **$6,256.01** (commit idle 16 h/day) |
| Jobs (no ingress) | same as instance-based | **$2,794.50** |